Research On Optimization Technology Of 3D Video Coding

Video,as an important means of disseminating information,has already become an indispensable part of people's daily life,work,study,entertainment,communication and so on.With the progress of science and technology,people's demand for video quality is also growing,which promotes the continuous change and development of video coding technology.Compared with traditional two-dimensional video,stereoscopic video can provide a more realistic and natural visual experience,and has a very broad application prospects.With the improvement of video resolution and the introduction of multi-view,network transmission bandwidth and processor computing capacity are facing more and more severe challenges.In order to standardize multi-view video coding and improve efficiency,a Joint Group of 3D Video Coding(JCT-3V)was established by the International Telecommunication Union(ITU-T)and the International Electrotechnical Commission(ISO/IEC),and a new generation of multi-view stereo video coding standard(3D-HEVC)was developed.Compared with the previous generation of multi-view stereo video coding standard(MVC),the average bit rate is reduced by about 50%.However,the computational complexity of the new generation of coding standards has increased exponentially,which makes it difficult to be applied directly.Therefore,it has great theoretical and practical significance to study the key technologies of 3D-HEVC and optimize its performance according to its mechanism.Under this background,this paper launches the optimization research of 3D-HEVC algorithm.Firstly,the dependence between texture/depth maps and synthetic viewpoints and the dependence between basic viewpoints and dependent viewpoints are analyzed in detail from the perspective of virtual view synthesis mechanism.On this basis,an optimized bit allocation scheme between texture/depth maps and basic/dependent view is proposed.Secondly,a novel rate control strategy with lower complexity for inter-view prediction is proposed,which not only studies the decision-making mechanism of initial quantization parameters,but also proposes an optimal bit allocation scheme for upper-level image groups and frames combined with buffer constraints.Thirdly,a rate-distortion optimization algorithm based on virtual view distortion estimation model is proposed to reduce the complexity of depth map coding.Finally,a hybrid depth map coding strategy is proposed by analyzing the influence of depth image elements on the quality of virtual view,and a balancing method between coding efficiency and coding quality is obtained.The proposed algorithm achieves rate control and high coding efficiency,and guarantees no buffer overflow in the coding process.On the premise of guaranteeing the quality of texture video,the complexity of depth map coding is greatly reduced and the coding efficiency is improved.The main innovative work in this paper includes:(1)Optimization of bit allocation technology based on synthesis view optimization Aiming at the joint coding resource allocation problem of multi-view texture video and corresponding depth map sequences,a solution based on virtual synthetic view distortion estimation model is proposed.Firstly,the rate-distortion model of multi-view texture video and the rate-distortion model of depth map sequence are modeled according to the relationship between the average distortion and the bit rate of texture video decoded and reconstructed.Secondly,the synthetic view distortion model is proposed and the virtual view-texture depth is obtained by combining the texture depth distortion model.Firstly,by studying the quality improvement of the base view to the dependent view,the view-dependent distortion model is deduced.Based on the above two distortion models,the joint rate allocation scheme of texture/depth and base/dependent view is constructed as an optimization problem under a constraint condition.Finally,Newton overlap is used to solve the problem.To solve and guide the resource allocation between viewpoints,we can maximize the quality of all viewpoints.(2)Rate control algorithm under buffer constraint conditions One of the important goals of rate control is to ensure that buffer constraints are met,that is,to ensure that no buffer overflow occurs in the encoding process,which requires rate control to adjust bit allocation according to the feedback of the buffer state in time during the encoding process,so as to generate the desired bit stream.Firstly,the original initial quantization parameter decision-making algorithm is analyzed.Based on the new feature of 3D video coding,an improved initial quantization parameter strategy and a rate control algorithm combined with inter-view dependence are proposed.This algorithm fully considers the quality-oriented smooth bit allocation scheme with buffer constraints,and achieves high coding efficiency while achieving rate control.(3)Depth map coding algorithm based on synthetic view distortion estimation Depth maps are used to guide the synthesis of virtual viewpoints rather than human eyes.In order to improve the coding efficiency of depth maps,it is necessary to obtain the synthetic viewpoint distortion caused by depth map distortion in the process of rate distortion optimization.In the current 3D-HEVC standard technology,the virtual view synthesis process is introduced into the depth map coding process to obtain the synthetic view distortion caused by the distortion of the current coded depth map.This process introduces huge coding overhead.In order to reduce the complexity of depth map coding,the mechanism of virtual viewpoint synthesis in 3D-HEVC is studied in this paper,and a synthetic viewpoint distortion estimation model based on interpolation is proposed to replace the high complexity of virtual viewpoint synthesis process to guide the rate-distortion optimization of depth map coding.(4)A mixed rate distortion optimization algorithm for low complexity depth maps Based on the analysis of the synthetic process of virtual viewpoint and the difference of the influence of texture and depth maps on the distortion of synthetic viewpoint under different occlusion relations,a depth image pixel classification method is proposed.The depth image coding unit(CTU)is divided into predictive coding CTU and base by this method.The two class of CTU is synthesized from the viewpoint,and then the mixing rate distortion optimization of depth map coding is performed.In this paper,the depth map hybrid rate-distortion optimization algorithm is proposed to reduce the complexity of depth map coding effectively while guaranteeing the coding quality.This paper combines the inter-view dependence with the rate control to realize the optimized bit allocation and accurate rate control.At the same time,it achieves more efficient depth map coding through the research of depth map and synthetic view optimization,which is a beneficial exploration for 3D video coding.